NFL preseason football kicks off this week, and three preseason weeks are about to produce more coverage than the games themselves merit. Twelve total games per team, most starters playing between 5 and 25 total snaps, and enough garbage-time production to inflate ten fantasy takes per rookie. The analytically useful signals inside all that noise are narrower than mainstream coverage suggests. Here’s what actually translates, what doesn’t, and how to read the coverage without absorbing the wrong lessons.
Quick read: NFL preseason analytics in 60 seconds
- What matters: Starter snap counts (coaching intent, expressed in reps), joint practice observations (real-game speed against opposing starters), specific rookie performance against starting units, and pressure rate in the limited snap samples where OL and DL starters actually face each other.
- What does not: Box-score statistics from bench-vs-bench stretches, single-play highlights extracted from ten hours of All-22, garbage-time drives run against defenders who won’t make the roster.
- The format trap: Most starters play 5-25 snaps across the entire preseason; extracting analytical inference from those samples is closer to reading tea leaves than reading data.
- What translates: Pressure rate against starting protection, route-running quality on limited film, defensive fit within the scheme, and the physical baseline questions (has the veteran lost a step?) that can be answered with the naked eye.
- What does not: Yards per attempt, completion percentage, total scoring volume — all of these are corrupted by which units are on the field and which aren’t.
What NFL preseason actually shows
The honest version of preseason coverage starts with a sample-size disclaimer that mainstream coverage rarely offers. A starting quarterback who plays 12 preseason snaps has, effectively, produced no data. A starting linebacker who plays two series against a mixed offense has produced no data. What preseason does show, in a way that matters, is coaching intent: which players are getting first-team reps in installation packages, which combinations the staff is testing in specific personnel groupings, and which battles have already been decided by the time preseason Week 2 arrives. That’s the layer that translates.
The vocabulary that supports this kind of preseason analysis lives in our sports analytics field guide, with the broader frame in our minicamps piece. Both treat preseason as a signal-extraction problem, not a prediction problem — which is the honest way to work with samples this small.
The preseason signal framework
Seven signals cover almost everything worth reading during preseason. Each has a defensible weight based on what the underlying sample can actually support.
| Signal | Weight | Why |
|---|---|---|
| Starter snap count distribution | High | Coaching staff intent, expressed in the language they trust most: reps |
| Joint practice observations | High | Real-game speed against another team’s starters — the best available preseason data by a wide margin |
| Rookie performance vs starters | Moderate-High | Genuine evaluation context; the rookie who holds up against a starting unit in preseason usually holds up in September |
| Pressure rate (limited samples) | Moderate | OL/DL battle baseline; even 15 snaps of starter-vs-starter reveals real matchup information |
| Box-score statistics | Low | Sample too small; scoring volume compounds coverage inflation |
| Backup quarterback performance | Low | Limited starter context; the backup who lights up preseason often faces defenders who won’t make Week 1 rosters |
| Garbage-time scoring | None | Bench-vs-bench production has been consistently uncorrelated with regular-season outcomes across every study |
A reading framework for preseason coverage
The seven questions below are the audit we run against any preseason coverage — team-specific beat reporting, national panels, fantasy content. Applied honestly, they filter out the majority of noise and surface the coverage worth reading.
| Question | What it reveals |
|---|---|
| How many snaps did the starter play? | Coaching staff confidence — a starter played more than expected suggests uncertainty in the depth chart below him |
| What did the joint practice reveal? | Real-game speed signals; joint practices are the closest thing to regular-season conditions preseason offers |
| How did the rookie perform against starters? | Genuine evaluation, once you filter for which starters (star vs backup-quality) |
| What does the second-string lineup look like? | Depth chart insight; the guys getting second-team reps in Week 2 are often the ones getting first-team reps in October due to injury |
| What injuries emerged? | Roster context that shapes every other reading — and often the biggest story of any preseason |
| How does the play-calling sample compare to camp? | Scheme installation progress; the offense running its full menu in preseason is different from the one saving it |
| What is the Week 1 outlook? | The actual question — the point of preseason is to inform the September answer, not to be its own answer |
The three preseason weeks, decoded
Each preseason week has its own signal profile, and treating them as interchangeable is a common analytical mistake. Week 1 is starter installation and rookie exposure; starters play 5-15 snaps, and the useful signal is which packages the coordinator is testing. Week 2 is the closest to a regular-season dress rehearsal; starters play into the second quarter, joint practices in the days before produce the best data of the preseason, and this is where the roster’s real depth chart becomes legible. Week 3 is bench-vs-bench roster battles for the last 10 roster spots; the signal here is almost entirely about the practice squad and 53rd-man decisions, not about the starting lineup.
Frequently asked questions
How predictive is preseason performance?
Modest at best. Starter snap counts and joint practice observations correlate with regular-season roles; box-score statistics from preseason games have consistently failed to predict September outcomes at rates better than chance.
What is the most underrated preseason signal?
Joint practices. Real-game speed against another team’s starters tells you more than preseason game snaps, because the reps are longer, the personnel matchups are more controlled, and the coaching staffs are actively working install against install. The problem is that joint-practice observations are underreported outside of local beat coverage.
Should I watch preseason games?
Yes if you want to see scheme installation and rookie physical baselines; no if you want to project regular-season performance from box scores. The games are useful as observation, not as prediction data.
Where can I read serious preseason coverage?
The Athletic’s team-specific writers for beat-level analysis, PFF for pressure-rate and rookie evaluation, and Pro Football Reference for the historical baseline of how preseason has (and hasn’t) translated in past years.
The takeaway, in one paragraph
NFL preseason produces narrow but real signals about scheme installation and rookie integration, buried under a lot of noise from bench-vs-bench box scores and single-play highlights. The framework above is the version we apply to any preseason coverage — and the discipline is being honest about which signals the underlying sample size can actually support. For the broader vocabulary, our sports analytics field guide is the natural companion read.
